THIS IS A TEST INSTANCE. ALL YOUR CHANGES WILL BE LOST!!!!
...
KStream involved in joins (including SStream-S Stream and SStream-TTable): if the KStream is ever generated from key changing operators, we cannot guarantee if it’s still partitioned by the key on its source topic, and hence need to repartition.
KStream involved in aggregates, following a groupBy function: as long as it is not groupByKey, we must repartition; and even if it is groupByKey, if the KStream has key changing operations preceded similar to 1), we also need to repartition.
KTable involved in aggregations, following a groupBy function: we will always repartition since we do not have KTable#groupByKey at all.
* Note that since there is no other key-changing operations for KTables except groupBy, KTable joins (TTable-TTable) never need to require repartitioning as its joining KTables are always safe to assume they are partitioned by key from their source topics (or from the previous repartition topics).
...